feat(tvix/store/blobservice): implement seek

For memory and sled, it's trivial, as we already have a Cursor<Vec<u8>>.
For gRPC, we simply reject going backwards, and skip n bytes for now.

Once the gRPC protocol gets support for offsets and verified streaming,
this can be improved.

Change-Id: I734066a514aed287ea3db64bfb1680911ac1eeb0
Reviewed-on: https://cl.tvl.fyi/c/depot/+/8885
Autosubmit: flokli <flokli@flokli.de>
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>
This commit is contained in:
Florian Klink 2023-06-30 16:08:58 +02:00 committed by clbot
parent 42dc18353d
commit 7613e2e769
6 changed files with 240 additions and 11 deletions

View file

@ -84,3 +84,135 @@ fn put_has_get(blob_service: impl BlobService, blob_contents: &[u8], blob_digest
assert_eq!(blob_contents, buf, "read blob contents must match");
}
/// Put a blob in the store, and seek inside it a bit.
#[test_case(gen_memory_blob_service(); "memory")]
#[test_case(gen_sled_blob_service(); "sled")]
fn put_seek(blob_service: impl BlobService) {
let mut w = blob_service.open_write();
io::copy(&mut io::Cursor::new(&fixtures::BLOB_B.to_vec()), &mut w).expect("copy must succeed");
w.close().expect("close must succeed");
// open a blob for reading
let mut r = blob_service
.open_read(&fixtures::BLOB_B_DIGEST)
.expect("open_read must succeed")
.expect("must be some");
let mut pos: u64 = 0;
// read the first 10 bytes, they must match the data in the fixture.
{
let mut buf = [0; 10];
r.read_exact(&mut buf).expect("must succeed");
assert_eq!(
&fixtures::BLOB_B[pos as usize..pos as usize + buf.len()],
buf,
"expected first 10 bytes to match"
);
pos += buf.len() as u64;
}
// seek by 0 bytes, using SeekFrom::Start.
let p = r
.seek(io::SeekFrom::Start(pos as u64))
.expect("must not fail");
assert_eq!(pos, p);
// read the next 10 bytes, they must match the data in the fixture.
{
let mut buf = [0; 10];
r.read_exact(&mut buf).expect("must succeed");
assert_eq!(
&fixtures::BLOB_B[pos as usize..pos as usize + buf.len()],
buf,
"expected data to match"
);
pos += buf.len() as u64;
}
// seek by 5 bytes, using SeekFrom::Start.
let p = r
.seek(io::SeekFrom::Start(pos as u64 + 5))
.expect("must not fail");
pos += 5;
assert_eq!(pos, p);
// read the next 10 bytes, they must match the data in the fixture.
{
let mut buf = [0; 10];
r.read_exact(&mut buf).expect("must succeed");
assert_eq!(
&fixtures::BLOB_B[pos as usize..pos as usize + buf.len()],
buf,
"expected data to match"
);
pos += buf.len() as u64;
}
// seek by 12345 bytes, using SeekFrom::
let p = r.seek(io::SeekFrom::Current(12345)).expect("must not fail");
pos += 12345;
assert_eq!(pos, p);
// read the next 10 bytes, they must match the data in the fixture.
{
let mut buf = [0; 10];
r.read_exact(&mut buf).expect("must succeed");
assert_eq!(
&fixtures::BLOB_B[pos as usize..pos as usize + buf.len()],
buf,
"expected data to match"
);
#[allow(unused_assignments)]
{
pos += buf.len() as u64;
}
}
// seeking to the end is okay…
let p = r
.seek(io::SeekFrom::Start(fixtures::BLOB_B.len() as u64))
.expect("must not fail");
pos = fixtures::BLOB_B.len() as u64;
assert_eq!(pos, p);
{
// but it returns no more data.
let mut buf: Vec<u8> = Vec::new();
r.read_to_end(&mut buf).expect("must not fail");
assert!(buf.is_empty(), "expected no more data to be read");
}
// seeking past the end…
match r.seek(io::SeekFrom::Start(fixtures::BLOB_B.len() as u64 + 1)) {
// should either be ok, but then return 0 bytes.
// this matches the behaviour or a Cursor<Vec<u8>>.
Ok(_pos) => {
let mut buf: Vec<u8> = Vec::new();
r.read_to_end(&mut buf).expect("must not fail");
assert!(buf.is_empty(), "expected no more data to be read");
}
// or not be okay.
Err(_) => {}
}
// TODO: this is only broken for the gRPC version
// We expect seeking backwards or relative to the end to fail.
// r.seek(io::SeekFrom::Current(-1))
// .expect_err("SeekFrom::Current(-1) expected to fail");
// r.seek(io::SeekFrom::Start(pos - 1))
// .expect_err("SeekFrom::Start(pos-1) expected to fail");
// r.seek(io::SeekFrom::End(0))
// .expect_err("SeekFrom::End(_) expected to fail");
}